452
Langmuir 1987, 3, 452-459
provided sufficient aging time was allowed. We believe
that these results will allow new insight into the time
dependence of the domain structure in a wide variety of
liquid-crystalline systems.
Acknowledgment. We thank R. M. Riddle (currently
of Texaco, Exploration and Production Research, Houston,
TX) for the construction of the multinuclear accessory for
the XL-100. This work was supported by the Department
of Energy and the National Science Foundation.
Registry No. SHBS, 67267-95-2; aerosol OT, 577-11-7.
Coadsorbate Interactions: Sulfur and CO on Ni( 100)
Marjanne C. Zonnevylle and Roald Hoffmann*
Department of Chemistry and Materials Science Center, Cornell University,
Ithaca, New York 14853
Received October 31, 1986. In Final Form: January 16, 1987
The influence of a S adlayer on CO adsorption onto Ni(100) is examined. Tight-binding extended Huckel
calculations on a three-layer model slab indicate that the interadsorbate separation distance determines
not only the mechanism but also the effect of the interaction. If the C-S distance is short, sulfur induces
site blockage of CO chemisorption by means of a direct, repulsive interadsorbate mechanism. If the separation
is increased beyond the normal S-C bond range, the sulfur adatoms work indirectly via modification of
the electronic structure of the substrate. This is a form of through-bond coupling. It is consistent with
the well-documented sulfur poisoning of CO adsorption and its usual explanation via relative electronegativities of adsorbates, but there are some conceptual differences. At longer coadsorbate separations,
there is an interesting reversal of the bonding trends, which has some experimental support.
The effect of atomic adsorbates on the chemisorption
of small molecules and on the rate of certain catalytic
reactions is dramatical Electronegative impurities, such
as halogens and chalcogens, tend to hinder these processes.
In contrast, alkali metals and other electropositive elements can function as promoters. A great variety of experimental and theoretical studies have been conducted
with the motivation of understanding the poisoning and
enhancement effects of coadsorbates. For example, the
adsorption of CO onto sulfided surfaces is often used as
a model for the sulfur poisoning of Fisher-Tropsch hydrocarbon catalysis from CO and H,. We will focus on
S/CO coadsorbates on nickel surfaces, as the body of experimental work dedicated to these systems is voluminous.
As is true of other coadsorbate systems, the basic nature
of the interactions between surface species is a matter of
controversy. Opinions are substantially polarized between
two extremes. The mechanism is described either as
dominated by delocalized long-range effects which allow
a single impurity adatom to modify many adsorption sites2
or alternatively as mediated by local bonding and shortrange site b l ~ c k a g e . ~Using the army of acronymical
methodologies available to surface science, experimental
evidence can be found to support either side. We list only
a few examples. Goodman and KiskinovaC4and Erley and
Wagner5 advocate the long-range theory based on the
well-documented nonlinearity of both the CO saturation
coverage and CO/H2 catalytic methanation as a function
of preadsorbed sulfur on Ni(100) and Ni(ll1). On Ni(100),
a one-fifth sulfur monolayer causes an order of magnitude
decrease in the rate of methanation. Goodmaneconcludes
(1) For recent reviews on poisoning and promotion, see: (a) Martin,
G. A. In Metal Support and Metal-Additiue Effects in Catalysis; Imelik,
B., et al., Eds.; Elsevier: Amsterdam, 1982; p 315. (b) Goodman, D. W.
"Role of Promoters and Poisons in CO Hydrogenation"In Proc. ZUCCP
Symp., TX, 1984.
(2) Goodman, D. W.; Kiskinova, M. Surf. Sci. 1981, 108, 64.
(3) Johnson, S.; Madix, R. J. Surf. Sci. 1981, 108, 77.
(4) Goodman, D. W.; Kiskinova, M. Surf. Sci. 1981, 105,L265.
(5) Erley, W.; Wagner, H. J. Catal. 1978, 53, 287.
(6) Goodman, D. W. Appl. Surf. Sci. 1983, 19, 1.
0743-7463 I 8 7 12403-0452$01.50 IO
that each sulfur atom affects some 10 Ni surface atoms.
Madix et al.3*798implicate local site blockage to explain
similar results. Gland et al.g favor the short-range model
for Ni(100) in light of high-resolution electron energy loss
spectroscopy (HREELS) and temperature-programed
desorption (TPD) studies. The infrared reflection-adsorption spectroscopy (IRAS) work of Trenary et al.1° on
Ni(ll1) is in good agreement with the latter.
Theoretical treatments of adatom poisoning and promotion of CO chemisorption are generally presented in the
framework of the Blyholder model." The adsorption
geometry is widely accepted to be through the carbon end,
exactly or nearly perpendicular to either the clean or
preadsorbed nickel surface.12"-d Regardless of the specific
adsorption site, the chemisorptive bond in the Blyholder
model results from the electron donation from the CO 5a,
la, into the empty surface levels and back-donation from
the surface into the CO 2 ~ * lb.
,
As both CO levels are
la
lb
20
2b
(7) Madix, R. J.; Thornburg, M.; Lee, S. B. Surf. Sci. 1983, 133, L447.
(8) Madix, R. J.; Lee, S. B.; Thornburg, M. J . Vac. Sci. Technol. A
1983, 1, 1254.
(9) Gland, J. L.; Madix, R. J.; McCabe, R. W.; DeMaggio, C. Surf. Sci.
1984, 143, 46.
(10) Trenary, M.; Uram, K. J.; Y a h , J. T.,Jr. Surf. Sci. 1985,157,512.
(11) Blyholder, G. J. Phys. Chem. 1964, 68, 2772.
(12) (a) Behm, R. J.; Ertl, G.; Penka, V. Surf. Sci. 1985,160, 387 and
references therein. (b) Allyn, C.; Gustafason,T.; Plummer, E. Solid State
Commun. 1978,28,85. (c) Andenson, S.; Pendry, J. B. Phys. Reu. Lett.
1979, 43, 363. (d) Passler, M.; Ignatiev, A.; Jona, F.; Jepsen, D. W.;
Marcus, P. M. Phys. Rev. Lett. 1979, 43, 360.
0 1987 American Chemical
Societv
Coadsorbate Interactions
localized on the carbon end,13they are at least geometric d y well suited for the electron transfers. Sulfur and other
electronegative adatoms withdraw electron density from
the surface. Little change in 5a to metal donation, 2a, is
presumed, but metal to 2ir* back-bonding, 2b, should be
reduced, and hence the poisoning effect. As the occupation
of the C-0 antibonding 2n* decreases, the C-0 bond
strengthens at the expense of the chemisorptive bond.
Electropositive adatoms should work in the opposite
manner. Electron density is pushed into the surface, thus
enhancing its back-bonding capacity.
Theoretical studies are able to reproduce poisoning and
promotion and most indicate, at minimum, a rudimentary
agreement with this simple MO picture. A cornucopia of
calculation methodologieshas been called upon, including
LCAO type on c l ~ s t e r s land
~ ~ - monolayer^,^^
~
monolayer
SLAPW type,16cluster LCGTO-XLU,"~~~
the effective-medium theory applied to jellium surfaces,lBa~b
the muffin-tin
approximation applied to clusters,lg and, most recently,
all-electron local-density-functional theory.20 Nethertheless, there is no more agreement (as to the range or
detailed nature of the interactions) between the various
calculations than that found in experimental studies.
Benzinger and Madix15suggest that direct interadsorbate
interactions may be the key factor, although the metalmediated electron transfers predicted from the Blyholder
mechanism are clearly reproduced. On the other hand,
Nerrskov et al.lEarbfind long-range electrostatic forces to
be more significant. The FLAPW results of Wimmer et
al.*O straddle the controversy. A long-range interaction is
dominant for a K adatom, but for S, both direct and
substrate-mediated effects are surmised.
In this paper, we examine the S + CO system on Ni(100)
with the extended Huckel tight-binding method.21a,b The
procedure is an approximate MO method with well-known
deficiencies. It does not give reliable potential energy
curves or binding energies, nor is it able to properly model
magnetism. The work function and ionization potentials
are quite far from the true values. But it is a transparent
methodology and reveals clearly the basic interactions that
are responsible for bonding. The calculations we report
are for nonmagnetic, spin-paired metal and adsorbate.
We will assemble the Ni/S/CO system from its various
component parts. Computational details are given in the
Appendix. As the focus of previous work by this group has
been the adsorption of CO onto Ni(100) and other surfaces,22several of the closely related systems-the bare
Ni(100) surface, the unadsorbed CO net, the Ni(100)-c(2
X 2) CO on top and Ni(ll1)CO bridging-are well characterized and our discussion will be appropriately brief.
(13)For the MO's of CO, see: Jorgensen, W. L.; Salem, L. The Organic Chemist's Book of Orbitals; Academic: New York, 1973;p 78.
(14)(a) Ray, N. K.; Anderson, A. B. Surf. Sci. 1983,125,803.(b) Ray,
N. K.; Anderson, A. B. Surf. Sci. 1982,119,35.(c) Anderson, A. B.; Surf.
Sci. 1977,62,119.(13)Tomanek, D.; Bennemann, K. H. Surf. Sci. 1983,
127,L111. (e) Cao, P.-L.; Shi, D.-H. Acta Phys. Sin. 1985,34,1291.(f)
Rodriguez, J. A.; Campbell, C. T., unpublished results.
(15)Benzinger, J.; Madix, R. J. Surf. Sci. 1980, 94, 119.
(16)Feibelman, P.J.; Hamann, D. R. Phys. Rev. Lett. 1984,52,61.
(17)(a) Dunlap, B.I.; Yu, H. L.; Antoniewin, P. R. Phys. Rev. A 1981,
25,7. (b) Jorg, H.; Rosch, N. Surf. Sci. 1985,163,L627.
(18)(a) Nsrskov, J. K.; Holloway, S.;Lang, N. D. Surf. Sci. 1984,137,
65. (b) Lang, N. D.; Holloway, S.;Nsrskov, J. K. Surf. Sci. 1985,150,24.
Pendrey, J.; Vvedensky, D. D. Surf. Sci. 1985,
(19)MacLaren, J. M.;
162, 322.
(20) Wimmer, E.; Fu, C. L.; Freeman, A. J. Phys. Reu. Lett. 1985,55,
2618.
(21)(a) Hoffmann, R. J . Chem. Phys. 1963,39,1397.Hoffmann, R.;
Lipscomb, W. M. Ibid. 1962,36,3179;Ibid. 1962,37,2872.(b) Ammeter,
J. H.; Biirgi, H.-B.; Thibeault, J. C.; Hoffmann, R. J. Am. Chem. SOC.
1978,100,3686.
i
(22) Sung, S.-S.; Hoffmann, R. J. Am. Chem. SOC.1985, 107,578.
Langmuir, Vol. 3, No. 4 , 1987 453
For the Ni(100) substrate, a three-layer slab with the
bulk nickel lattice constant ( a = 3.524 AZ3)was used. The
nearest-neighbor contact is 2.49 A on this square net
surface. The calculated density of states (DOS) exhibits
a compact d band between -12 and -8 eV and very disperse
s and p bands between -12 and 8 eV.22,24As the surface
atoms have a lower coordination number, these states are
less disperse than the bulk states. Consequently, if the
Fermi level falls above the midpoint of the d block, as for
nickel, the surface is negatively charged with respect to
the bulk.
Whether the three-layer slab is an appropriate model
for the surface can be estimated from the charge on the
middle layer. An infinitely deep system will have a bulk
charge approaching zero. With a 55K point set,25we
compute an excess charge of +0.180e- per atom for three
layers, +0.166e- for four layers, and +0.157e- (middle layer)
for five. As the convergence is slow in this range of reasonably sized unit cells, we make our choice on the basis
of computational economics. The three-layer substrate was
used for all calculations.
Our first priority in the study of the S/CO system on
Ni(100) is to establish proper adsorption geometries. So
let us review briefly what is known from experiment. The
preference for atomic adsorption at the highest coordination site is nearly universal on clean or coadsorbed transition-metal surfaces.% Specifically, chalcogen adsorption
at the Ni(100) four-fold hollow is well d o c ~ m e n t e d . ~ ' ~ - ~
CO chemisorbs molecularly and does not dissociate to any
measurable degree on late-transition-metal surfaces.28 On
clean Ni(100), CO site preference is coverage and temperature dependent.BH At saturation coverage and below
150 K, a compression structure of both on-top and bridging
CO is formed. When the compression structure is relaxed
to a 6 = 0.5,c(2 X 2) monolayer (or if coverage is further
reduced), only the on-top mode is observed. The HREELS
vibrational spectrumg is characterized by a single C-0
stretching frequency at 2005 cm-'. If sulfur is added, and
its surface concentration is increased, this dominant peak
is down-shifted and two new CO modes grow in, one at a
lower frequency and the other at a higher frequency. The
adsorption geometry of the new modes is unknown. Using
partial thermal desorption, the high-frequency peak is
correlated to low-temperature desorption, the low-frequency peak to high temperature, and the dominant mode
to an intermediate temperature. The CO layer is completely desorbed below 430 K, which is the predomiant
desorption temperature from the clean surface. The inverse relationship between the C-O stretch frequency and
desorption temperature is consistent with the greater
susceptibility of the chemisorptive bond strength to the
n rather than the u system, as advocated in the Blyholder
mechanism. However, we expect a unidirectional effect;
(23)Donohue, J. The Structure of the Elements; R. E. Kreiger: Malabar, FL, 1982.
(24)Saillard, J.-Y.;Hoffmann, R. J.Am. Chem. SOC.1984,106,2006.
(25) Pack, J. D.; Monkhorst, H. J. Phys. Reu. B 1977,16,1748.
(26)For example, we: van Hove, M. A. In The Nature of the Surface
Chemical Bond; Rhodin, T. N.; Ertl, G., Eds.; North-Holland: Amsterdam, 1979; p 277.
(27)(a) Fisher, G. B. Surf. Sci. 1977,62,31.(b) Perdereau, M.; Oudar,
J. Surf. Sci. 1970,20,80.(c) Demuth, J. E.; Jepsen, D. W.; Marcus, P.
M. Phys. Rev. Lett. 1973,31,540. (d) Andersson, S.Surf. Sci. 1979,79,
385. (e) Hagstrum, H. D.; Becker, G. E. J. Chem. Phys. 1971,54,1015;
Phys. Reu. Lett. 1969,22, 1054.
(28) BrodBn, G.; Rhodin, T. N.; Brucker, C. F.; Benbow, R.; Hurych,
Z. Surf. Sci. 1976,59,593.
(29)(a) Andersson, S. Solid State Commun. 1977,21, 75. (b) Andersson, S. In Proceedings of the 7th International Vacuum Congress
& 3rd International Conference on Solid Surfaces;Vienna, 1977;p 1019.
(c) Mitchell, G. E.; Gland, J. L.; White, J. M.Surf. Sci. 1977,62,31.
454 Langmuir, Vol. 3, No. 4, 1987
Zonneuylle and Hoffmann
Total DO8
NI rurfacs XZ+YZ
_ _ _ _ - _ _ - - - - - -x t y
s
- s 3p
-26
Nl(100) DO5
co
8
Figure 1. In the left panel, the bare Ni(100) DOS. The valence
orbitals of CO and S are drawn in the middle and right panels,
respectively.
the sulfur should reduce r-back-bonding. The C-0 bond
strengthens as the chemisorptive bond weakens, thereby
creating a high CO frequency, low desorption temperature
mode. Although two such modes are observed, a third,
apparently originating from increased a-back-bonding,
does as well. We ambitiously hope to model these results
by examining the effect of sulfur coadsorption on the
binding of CO at various sites and, as well, gain general
insight into the mechanism of coadsorbate interactions.
In the case of the adsorbed systems, we retained the Ni
bulk geometry for the three-layer substrate slab. (2 X 2)
chalcogenide overlayers do not cause reconstruction of this
surface; however, relaxation must be considered. For the
clean Ni(100) surface, the spacing between the first and
second layers is estimated to be contracted by 3 % , based
on ion blocking m e a s ~ r e m e n t s . ~EELS
~
studies indicate
that no further relaxation occurs for (2 X 2) S ad layer^.^^
To our knowledge, it is not known whether reconstruction
or relaxation occur in the presence of both S and CO.
While the geometric changes are not insignificant, they are
not included in our calculations.
A coverage of 9 = 0.25, p(2 X 2), per adsorbate species
was chosen. It is in this coverage range that the three (3-0
stretch peaks first appear simultaneously in the HREELS
~ p e c t r a .Each
~ species is laid out in a square net, lattice
constant a = 4.98 A. The calculated band structures of
COZ2and S unsupported nets are essentially free of dispersion. The band energies are equivalent to the molecular
orbital energies, thus intraspecies interactions are negligible
at this distance. To help us see the important orbitals of
both adsorbates and their relative energies, we show in
Figure 1the total density of states (DOS) of the bare Ni
slab and, alongside, the important frontier orbitals of S
and CO.
One of the prerequisites for the electronegativity arguments concerning poisoning is that the electron-withdrawing power of sulfur be observed in Ni(100)-p(2 X 2)
S, that is, without CO coadsorption. K point sets of 10
(30) Frenken, J. W. M.; Smeek, R. G.; van der Veen, J. F. Surf. Sci.
1983,135,147. Frenken, J. W. M.; van der Veen, J. F.; Allen, G . Phys.
Reu. Lett. 1983,51, 1876.
(31) Lehwald, S.; Rocca, M.; Ibach, H.; Rahman, T. S. J. Electron
Spectrosc. Relat. Phenom. 1986, 38, 29.
- s 38
DOS
Figure 2. Solid line indicates the Ni(100)-p(2
X 2) S total DOS.
The dotted and dashed linea are the integrals of the surface xz
+ yz and S x + y projected DOS,respectively. The S unsupported
square net DOS is represented by the median energy bars on the
right.
and 16 were used for substrate plus adsorbate(s) systems
with tetragonal and orthorhombic symmetry, respectively
(see Appendix for calculations testing convergence). The
sulfur square net is placed on the substrate such that S
occupies 4-fold hollow sites and d(S-Ni) = 2.19 A,32as
obtained from LEED. The z axis is consistently oriented
perpendicular to the Ni plane, and the x and y axes are
parallel to the surface Ni-Ni close contacts. The charge
transfer occurs as predicted; the sulfur adatoms gain
0.756e-, and each of the four associated surface atoms loses
0.355e- relative to bare Ni(100). Although the charge
transfer is greatest in the nickel x z and y z (-0.115e- each)
and the sulfur z levels (+0.481e-33), the projected DOS
indicate that the strongest interaction is with the sulfur
x and y rather than z levels. The p band of the S square
net lies at -13.2 eV, well below both the d block and the
Fermi energy of the bare substrate (-8.7 eV). The large
dispersion provided by the interactions with surface xz and
yz states, such as the representative one depicted in 3,
(32) Reference 26, p 292.
(33) The reference occupation for sulfur atomic orbitals is 2e- for s and
1.333e- for each p.
Langmuir, Vol. 3, No. 4 , 1987 455
Coadsorbate Interactions
forces some 20% of the S x and y states above the Fermi
energy and splits the major S x and y peak down from the
z peak by -0.5 eV. Some surface x z and yz character
appears in the S x and y peak near -14 eV, as seen in a
comparison of the projected DOS integration curves in
Figure 2. In contrast, the S z level acts essentially as an
inert electron sink.
The total DOS (solid line in Figure 2) appears very much
as a simple overlay of the S square net DOS (represented
schematically on the right by bars at the median energies)
and the bare Ni(100) DOS.26 The binding energies of the
sulfur p and s levels agree with those obtained by UPS2’*
within experimental error. The S adlayer is a relatively
small perturbation on the electronic structure of the Ni(100) slab as a whole but does cause substantial charge
redistribution and rehybridization at a local level. If
poisoning does occur via the substrate, the x z and yz
surface levels are the most likely mediators.
We consider three high-symmetry sites for CO coordination: on top, bridging, and four-fold hollow. The C-Ni
and C-0 bond lengths are determined by LEED as 1.80
and 1.15 A, respectively,12 for on-top coordination, the only
one known to exist independently. In general, our computational method cannot be trusted to predict the energetics of adsorbate site preference. So we must make
some geometric choices. For the bridging and 4-fold sites,
all calculations on both clean and sulfur-coadsorbed surfaces were performed at two limiting geometries: C-Ni
nearest-neighbor distance of 1.80 A and C-Ni surface
distance of 1.80 A. The trends are identical but consistently more pronounced for the first option, which we
choose to present for the sake of clarity.
Characteristic of the rather weak chemisorption of CO
onto nickel and other late transition metals, the CO molecular levels for the observed on-top geometry form bands
which have a small dispersion and are at nearly the same
energy as the molecular levels. In contrast, one-third of
the bridging and 4-fold 2n* levels are pulled into the d
block. Specific features of the DOS will be discussed later,
in conjunction with the coadsorbed systems.
Our models adhere closely to the Blyholder model of
a-donation and n-back-donation. The electron density
shifts from bare Ni(100) or molecular CO to Ni(100)-p(2
x 2) CO are presented in Table I. The extent of 5a
depopulation falls within a much narrower range than does
the amount of back-donation into 27r*.34a,bThe a interaction contributes to the net bonding in a relatively uniform manner for all sites. On the other hand, the C-0
overlap population (op) follows the 2a* occupation exactly
as predicted. Moving through the coordination series of
molecular, on top, bridging and 4-fold, the C-0 op falls
in conjunction with growing 2n* occupation. The CO
coordination number mediates the electronic shifts at the
associated surface atoms, the largest shift coincides with
the lowest coordination. The average electron loss, however, correlates well with the gain in 2n* but is independent
of 5a. These two effects underlie the notion that the
chemisorptive interaction is governed by the extent of
n-back-bonding against a fairly uniform a-donation
background. Unfortunately, we cannot compare the
strengths of the chemisorptive bonds directly; the effect
of coordination number on overlap population is generally
nonlinear.
(34) (a) Bagus, P. S.; Nelin, C. J.; Bauschlicher, C. W., Jr. J. Vacuum
Sci. Technol. A 1984,2,905. (b) Bauschlicher, C.W., Jr.; Bagus, P. S.
J.Chern. Phys. 1984,81,5889.These workers seem to conclude that the
net u interaction contributes little to the chemisorptive bond, with which
we are not in agreement.
Table I. Calculated Results for Ni(100)-p(2 X 2) C O
CO coordination mode
on top
4-fold
bridging
atom(s) or
orbitals
Electron
CO total
co 5a
2T*b
Ni,, coordinated
Ni,, uncoordinated
Ni,, average
Density Changes”
+0.249
+0.777
-0.368
-0.379
+0.372
+0.694
-0.283
-0.814
-0.005
-0.283
-0.200
+0.638
-0.398
+0.598
-0.542
-0.002
-0.273
Overlap Populations
1.044
0.916
0.847
0.359
0.942
0.631
co
c-oc
C-Ni8
CO binding E, eVd
+2.570
+2.526
+3.843
“Relative to Ni(100) or molecular CO. Net electron gain if positive, loss if negative. *Occupation of 27r* and other degenerate
orbitals are given for each individual orbital. ‘Molecular C-0 op
= 1.208. dE(Ni(lOO)) + E(molecu1ar CO) - E(Ni(100)-p(2 + 2)
CO).
Atomic electron transfers must, however, be used with
care. In this case, both CO interactions work to decrease
the surface electron density. Most of the d block lies just
below the Fermi energy so that some states will be pushed
above it by dispersion from any source, whether from
below (5a) or above ( 2 ~ * ) That
. ~ ~ d-block states can be
raised above ef by interaction with 2n* merits a word of
discussion. Very simplistically, mixing between a fully
occupied “d” and empty =2n*” states will create a “d +
X2n*” state below ef and a “2n* - Ad” state above ef (A is
an arbitrary mixing parameter). The net charge transfer
will be out of d and into 2a*. Although the bare-surface
d-block occupation is actually 9.3e- rather than 10.0e-, it
is not hard to imagine that simply by their preponderance,
the occupied states will dominate. The two effects are
most cleanly separated at the on-top geometry. The orbitals able to interact with the 5a do not interact effectively
with the 2n*, and vice versa. The appropriate surface d
orbitals (z2 for 5a, x z and yz for 2n*) experience nearly
identical charge shifts. Since the 5a occupation is very
similar at the three sites, we can safely surmise that its
interaction with the substrate is constant and the total
surface electron loss is determined by the 2n*.
With regard to the electronegativity arguments concerning sulfur poisoning, we suggest that the electronwithdrawing power of CO must be considered as well. The
total charge transfer into CO is approximately one-third
of that into the S of Ni(100)-p(2 X 2) S for CO on top but
nearly equivalent for the bridging and 4-fold coordinations.
Although the point must be reconsidered for the coadsorbed systems, this result makes rationalizations based
on electronegativity alone less self-evident.
Four Ni(100)-p(2 X 2) S + p(2 X 2)CO systems have
been considered (1)CO on top, 4a,(2) CO in 4-fold hollow
retaining tetragonal symmetry (4-fold tetra), 4b; (3) CO
in 4-fold hollow, symmetry reduced to orthorhombic (4-fold
ortho), 4c; and (4) CO bridging, 4d. The top layer of the
three-layer unit cell is shown in the schematics. In each
case, we have left the S in a four-fold hollow. Behind this
assumption is the strong propensity of the S for this site;
to our knowledge, no deviation from such adsorption is
known experimentally. CO, on the other hand, is known
to have low barriers to moving into alternative coordination
modes.
The evolution of the C-0 and C-Ni, op’s (Ni, = surface
atom) upon sulfur coadsorption, shown in Table 11, is intriguing in light of the Blyholder and “anti-Blyholder”
modes observed by Gland.g Just as predicted by the
Blyholder mechanism, S coadsorption in the 4-fold ortho,
Zonnevylle and Hoffmann
456 Langmuir, Vol. 3, No. 4 , 1987
0
Table 11. Overlap Populations and Binding Energies
C-0 overlap
C-Ni8 overlap
n
DoDulation
without with
4a
4b
0
Doaulation
without with
CO site
S
S
S
S
ACO binding
E, eVa
on top, 4a
4-fold tetra, 4b
4-fold ortho, 4c
bridging, 4d
1.044
0.916
0.916
0.942
0.929
0.906
0.928
0.974
0.847
0.359
0.359
0.631
0.579
0.363
0.336
0.581
+3.391
-0.050
+2.435
+1.256
/A*j7
(CO binding E with
S)- (CO binding E without S).
Table 111. Occupation of 2?r*
4c
4d
4c, and bridging, 4d, systems induces C-0
bond
strengthening and C-Ni, bond weakening and reduces the
CO binding energy. Although the absolute binding energies calculated by the extended Huckel method are unreliable and we are unable to predict adsorption site
preferences, the relative binding energies a t the same site
can be used with more confidence. We tentatively associate the 4-fold ortho and bridging sites with the highfrequency C-0 stretch, low-temperature desorption mode.
Exactly the opposite effects are observed for the 4-fold
tetra, 4b. This site corresponds to the "anti-Blyholder"
low-frequency C-0 stretch, high-desorption-temperture
mode. That the interaction between adsorbates is nonisotopic is not without precedence. Other w o r k e r ~ have
~~q~
proposed that the interaction is an oscillatory function of
the separation distance, thus alternately repulsive or attractive.
Both bonds are weakened if CO is bound on top, and
the CO binding energy decreases so sharply that the Ni(100)-p(2 x 2) S + p(2 X 2) CO system is less stable than
its separated components. The anomalous behavior arises
directly from the close C-S contact. In this geometry, 4a,
with 1/4 coverage for both S and CO, the S-C(O) separation
is only 1.83 A. This is very much within a bonding range;
indeed typical S-C single bonds fall between 1.75 and 1.80
A. The computed C-S op is 0.627, which is the largest
value of any site (the bridging mode is a distant second
at 0.015). The surface species is better described as molecualr SCO than as coadsorbed S and CO. To our
knowledge, no evidence of SCO formation has been observed for any coadsorbed S / C O surface. The energy of
this coadsorbate geometry is very high. We believe it is
energetically inaccessible and that what we are in fact
modeling is local site blockage.
Essentially all other calculations to date have been
preformed at similarly short C-S contacts. This distance
is crucial; at 1.83 A we observe site blockage, at 2.66 A
(4-fold ortho) and 2.78 A (bridging) the Blyholder result,
and at 3.64 A (Cfold tetra) an "anti-Blyholder" effect. The
smallest distance must be classified as short-range and the
large S-C op strongly indicates a direct interadsorbate
effect; the largest is clearly long-range and its corresponding op is negligible (-0.003), pointing to a surfacemediated interaction. The problem, then, may not be to
determine the absolute range of interadsorbate interactions
but rather to identify the type of interaction active in each
range.
An important point in our analysis is that the 2a* occupation follows the strength of the chemisorptive bond.
The trends shown in Table I11 are consistent with the
rr-back-bonding mechanism; invariably C-Ni, op reduction
is coincident with 2a* depopulation, and vice versa.
The relative electron-withdrawing power of the two
adsorbates is reconsidered in Table IV. Relative charge
transfer is defined as the atomic electron density in Ni-
predicted
change"
-
CO site
on top
4-fold tetra
4-fold ortho
bridging
occupation
without
with
S
0.372
0.694
0.694
0.598
+
-
S
0.515
0.717
0.686
0.521
'Based on C-NI overlap population trends.
Table IV. Electron Transfer to Coadsorbates Relative to
Single-Species Systems"
CO site
S
co
on top, 4a
-0.530
-0.009
4-fold tetra, 4b
4-fold ortho, 4c
bridging, 4d
+0.015
-0.041
-0.051
+0.071
-0.050
-0.179
a A positive value indicates an electron density gain relative to
single-species adsorption and a negative value, an electron loss.
Table V. Electron Density Changes,"Bridging CO
surface
atom'
-0.542 (-0.002)
Ni(100)-p(2 X 2) S + p(2 X 2)
CO re1 to
Ni(100)-p(2 X Ni(100)-p(2 x
2) co
2) s
-0.179
-0.022
-0.144
-0.037
-0.046
-0.028
-0.376 (-0.263) -0.563 (-0.090)
S
+0.004
-0.188
-0.262
+0.010
-0.120
-0.024
-0.190
-0.109
Ni(100)-p(2 X 2)
CO re1 to
atom(s) or
orbitals
Ni(100) or CO
CO total
+0.638
co 5rJ
-0.398
CO 2 ~ * , ~ +0.746
Co 2a*,
+0.449
S total
s x,y
22
XZb
YZ
(-0.005)
(+0.006)
(-0.087)
(-0.024)
(-0.034) -0.039
(-0.065) -0.094
(-0,091) -0.337
(-0.089) +0.016
(-0.001)
(+0.011)
(-0.063)
(+0.002)
a Net electron gain if positive, loss if negative. *Bridging direction along x. CBridged(unbridged).
(100)-p(2 X 2) S + p(2 X 2) CO minus that in the single
adsorbate system (Ni(100)-p(2 X 2) S or Ni(100)-p(2 X
2) CO). A positive value indicates that the coadsorbed
species acquires electron density; a negative value indicates
a net loss. According to this indicator, the S is more
electronegative than 4-fold ortho or bridging CO but more
electropositive than 4-fold tetra CO. The results are
consistent with the notion that a more electronegative
species, the 4-fold CO tetra, will accept more back-bonding.
Although the results thus far are consistent for each site,
the question remains as to why these sites behave in fundamentally different ways. Let us begin with the bridging
geometry. The arguments for the 4-fold ortho site are
much the same.
The composite characteristics of the Ni(100)-p(2 X 2)
S + p(2 X 2) CO bridging total DOS can be seen clearly
in Figure 3. The DOS of the CO and S unsupported nets
are represented by bars at the right. In spite of energy
Langmuir, Vol. 3, No. 4, 1987 457
Coadsorbate Interactions
0 ,
.
I
'
co 27-l'
4
C04u
DOS
Figure 3. Bridging Ni(100)-p(2 X 2) S + p(2 X 2) CO total DOS;
major peaks are labeled. The bars on the right indicated the
median energies of the CO and S unsupported nets.
shifts, their peakedness is substantially retained; a minimum of 50% of the adsorbate levels fall in their major
peaks as labeled. As previously discussed, the direct interaction is predictably small, with d(C-S) = 2.78 A and
C-S op = 0.015. Some mixing occurs between the CO l a
and the sulfur p levels since they are energetically closely
matched. But the la is heavily localized at the oxygen end
(see Figure 1)and is therefore relatively inert to adsorption.
The poisoning must occur via the substrate.
The same conclusions can be drawn from the electron
density shifts, of which a selected few are listed in Table
V. The excess charge at S is nearly independent of CO
adsorption. In contrast, the local surface charge is very
much determined by the adjacent surface species. With
only one exception, every atom of the initially negatively
charged surface (bare Ni(100)) loses electron density as
adsorbates are added sequentially. The bridged atoms of
the coadsorbed system suffer most with respect to either
single species slab.
Let us focus on the x z orbitals, which lie along the
bridging direction. As discussed earlier, the electronwithdrawing power of S is felt most strongly at this orbital
and its degenerate partner in Ni(100)-p(2 X 2) S. The
largest perturbation of CO adsorption onto either the clean
or sulfided surface occurs here as well. The x z is geometrically well suited for interaction with both 5a and the
2a*,, which lies along the bridging direction, depending
on the phase relationship between the bridged atoms. The
out-of-plane combination can participate in a-bonding, 5a,
and the in-phase in a-back-bonding, 5b. The interactions
are energetically favored as well. The out-of-phase states
will dominate the lower half of the d band (near 5a) since
they are bonding between the metal centers. Likewise, the
in-phase set will be concentrated in the upper half, adjacent to the 2a*. The interaction with the 2 ~ * is
, particularly strong; the occupations of the originally degenerate
Sb
Sa
molecular orbitals differ by 0.297e- and the median energies are split by -le V. Later, we will use the phase
relationships between the Ni atoms and a technique we
call COOP to identify the a and a interacting states.
The stepwise evolution of the x z can be traced in Figure
4. The integral of the projected DOS is compared for
Ni(100), Ni(100)-p(2 X 2) CO, and Ni(100)-p(2 X 2) S +
p(2 X 2) CO and overlayed onto the total DOS of the latter.
The initial localization of the states entirely within the d
block is disrupted by both u and a effects. The mixing
is evident from the substantial redistribution of levels into
both the 5a and 2x* areas. Both effects are magnified with
addition of the S adlayer.
If the integrals of the 5u projected DOS are compared
for the surfaces with and without S, we can see that sulfur
addition pulls some of the 5a states into the CO 1~and
sulfur p peaks but does not substantially alter either the
median of energy or the distribution of levels near q. On
the other hand, a similar comparison of the 2 ~ * integrals
,
reveals that the 2x*, level is strongly distorted and the
median is pushed up -le V. We must remember, however, that these effects cannot be caused by a direct S-CO
interaction.
How does the depopulation of the CO 2n* come about
on the sulfided surface? We saw earlier that the 2a* interacts strongly with the upper part of the metal xz band
(5b). The S p orbitals also interact with the same part of
the band and push it up (Figure 4). The metal-2a* interaction becomes stronger since the energy difference is
reduced. As a result, both the adsorbate and metal levels
are depopulated. To explain why, we turn to the simple
two-level interaction diagram used so commonly for molecular systems, and modify it to indicate the inherent level
dispersion of infinite systems, 6. The schematic is crude,
perhaps, but functional nonetheless. The heightened interaction naturally forces upward the resultant antibonding
levels, which contain both metal and 2 ~ character.
*
Contrary to molecular systems for which the energy of the
HOMO can change due to interactions, the fermi level of
the metal surface is essentially fixed, regardless of the
surface species. Our three-layer calculation models this
well; the greatest shift of tf after adsorption is 0.13 eV from
the value of the bare surface. The placement of ef in 6 is
crucial. In this case, the arrangement is such that many
of the levels are pushed above E~ A similar interaction in
molecular systems will bring about a gain in the occupation
of one of the levels at the expense of the other, but here,
because cf is fixed, both metal x z and CO 2x* levels lose.
An accompanying diagram could be drawn for the CO 5a
interaction with the bottom of the x z band (5b). However,
most of the levels will lie much below eft so that even an
interaction of similar magnitude will not alter the 5a oc-
458 Langmuir, Vol. 3, No. 4 , 1987
Zonnevylle and Hoffmann
NI-NI, wlth S & CO
NI-NI. wlth CO
Nl(lOO)-p(2~2)S+p(2~2)CO
xz, wlth S & CO
xz, wlth CO
...... - - - xz, bare rurface
__I___
.................
0
............
.I......::::.
-4......
....
.:...
.....
.d.
i
ff::
.........:::............::.:::::::.
.....
0.
co 2lT*
- co 27r*
-
-10
-26
r-
.............
.........
.....
co 5v
1
-26
tAntlbondlnp
Bonding-+
Figure 5. Bridged Ni-Ni COOP’S for the Ni(100)-p(2 X 2) CO
and Ni(100)-p(2 X 2) S + p(2 X 2) CO are indicated by solid and
r
dotted lines, respectively.
Table VI. Electron Density Changes,” 4-Fold Tetra CO
DOS
Figure 4. Dashed, dot-dashed,and dotted lines are respectively
the integrals of the x z projected DOS of Ni(100), Ni(lOO)-p(2 X
2) CO, and Ni(100)-p(2 X 2) S + p(2 X 2) CO. The solid curve
is the total DOS of the latter. The median energies of the 5u and
2n* coadsorbed levels are indicated on the right.
cupation to the extent of the 2+.
Should this interaction, by which S modifies the ability
of some surface-localized metal states to bond to the 27r*
of CO, be called a direct one? We would prefer to refer
to it as “through-bond coupling”, a concept that has served
well in ~ r g a n cand
~ ~inorganicSasb
~,~
chemistry. Seemingly
localized orbitals (lone pairs in’two parts of an organic
molecule or two metal d orbitals in a binuclear complex
connected by one- or many-atom bridges) in two or more
parts of a molecule can mix, interact, and split in energy
as a result of mixing with orbitals of intervening atoms or
groupings of atoms. This is what happens here.
That the 5u interaction is modified at all can best be
followed indirectly through the crystal orbital overlap
population (COOP) curve37of the bridged nickel atoms.
The COOP is created by weighing the DOS by its contribution to the Ni-Ni op. In Figure 5, these are compared
for t,he clean and sulfided systems. From 5a and 5b, we
know that some Ni-Ni bonding states (near the bottom
of the d band) can interact with the 50. and that some
.
the resantibonding ones interact with the 2 ~ * Indeed,
(35) (a) Grimley, T. B. Pontif. Acad. Sci., Scr. Varia 1976, 31, 443;
Proc. Phys. SOC.(London) 1967,90, 751. (b) Einstein, T. L.; Schrieffer,
J. R. Phys. Reu. B 1973, 7, 3629.
(36) (a) Hoffmann, R. Acc. Chem. Res. 1971, 4, 1. (b) Gleiter, R.
Angew. Chem. 1974,86,770; Angew. Chem., Znt. Ed. Engl. 1974,13,696.
(37) (a) Shaik, S.; Hoffmann, R.; Fisel, C. R.; Summerville, R. H. J.
Am. Chem. SOC.1980,102,4555. (b) Whangbo, M.-D. In Crystal Chemistry and Properties of Materials with Quasi-One-DimensionaZStructures; Rouxel, J., Ed.; D. Reidel: Dordrecht, Holland, 1986; pp 27-85.
(38) Hughbanks, T.; Hoffmann, R. J . Am. Chem. SOC.1983,105,1150.
Wijeyesekera, S.; Hoffmann, R. Organometallics 1984, 3, 949.
atom(s) or
orbitals
CO total
co 5u
co 2n*
S total
s X,Y
Ni(100)-p(2 X 2)
CO re1 to
Ni(100) or CO
+0.777
-0.368
+0.694
-0.283
22
-0.043
-0.133
+0.003
.=,
a
YZ
+0.064
+0.009
+0.023
+0.015
+0.005
surface atom
XY
Ni(100)p(2 X 2) S + p(2 X 2)
CO re1 to
Ni(100)Ni (100)p ( 2 X 2)
p(2 x 2)
CO
Ni(100)
S
-0.338
-0.071
-0.023
-0.129
-0.621
-0.114
-0.154
-0.126
-0.266
-0.044
-0.132
-0.011
Net electron gain if positive, loss if negative.
onances with the CO are picked up and the shift from
bonding to antibonding through the d block occurs as
expected. However, a small bonding region (circled in
Figure 5) appears above ef of the sulfided slab. Assuming
conservation of the absolute number of bonding and antibonding states, this packet must have been pushed up
from the d block by the sulfur and again by the CO 5u.
Surface levels that should have been filled by u-donation
remain empty. The u contribution to the chemisorptive
bond must be reduced, regardless of the stability of the
5u occupation. Since the number of affected levels is not
substantial, the 2ir* occupation remains the overriding
factor.
Can analogous arguments be used to explain the antiBlyholder behavior of the 4-fold tetra system? The charge
densities computed for the coadsorbed slab lie much closer
to the single species values than do their bridging counterparts. A few are listed in Table VI. That the magnitudes are small should not concern us unduly as the
trends tend to be more reliable in similar extended Huckel
calculations. Moreover, the trends are consistent among
themselves. For example the 27r* occupation remains
Langmuir, Vol. 3, No. 4, 1987 459
Coadsorbate Interactions
Table VII. Extended Huckel Parameters
orbital
Hii, eV
orbital
Hii, eV
2.27
0 2s
-29.6
-7.8
2.1
Ni 4s
4p
3d4
C 2s
2p
-3.7
-9.9
-18.2
-9.5
2.1
5.75
1.63
1.63
2p
S 35
3p
-13.6
-20.0
-13.3
2.27
1.817
1.817
= 2.00; Cl = 0.5683; C2 = 0.6292. C = contraction coefficients used in double-{ expansion.
directly proportional to the C-Ni, op and indirectly proportional to the C-0 op.
The Cfold tetra system differs from the others in several
respects. I t is unique among the coadsorbed systems in
that S and CO are each able to withdraw more electron
density from the substrate than if adsorbed individually.
It is also the only system for which the electronic shifts
in the surface levels caused by the coadsorbates are nearly
additive. The sum of the charge transfer between the bare
and single adsorbate Ni(100) surfaces (columns 2 plus 4
of Table VI) is nearly equivalent to the shifts between the
bare and coadsorbed values (column 3). The largest difference is in the degenerate x z and yz orbitals. Individually, S and CO remove a total of 0.122e- from each, but
if coadsorbed, 0.014e- more. As a comparison, the greatest
discrepancy for the bridged CO systems is nearly 5 times
the value (in z2). In addition, the coadsorbates work on
entirely different parts of the surface. This can be seen
by comparing the relative charge transfers between coadsorbed and single-species slabs to coadsorbed and bare
slabs (columns 2-5 in Table VI). The effect of CO on the
clean or sulfided substrate is greatest at the surface xy and
smallest at the x z and yz. The reverse is observed for the
sulfur modification of the clean or CO-preadsorbed slabs.
Thus the adsorption characteristics of either species are
nearly independent of the perturbation of surface levels
caused by the other. The same information can be obtained by overlaying the appropriate projected DOS curves
and integrals.
The answer may lie in the middle layer, rather than the
surface. Adsorption of either S, CO, or both will increase
the electron density of the bulk layer relative to that of
the bare slab. The effect of approximately additive (Le,,
the sum of the single species equals coadsorbed system)
in every case save the 4-fold tetra site. Here, the electron
transfer is slightly more than half as much as expected.
A comparison of the bulk atom projected DOS curves
uncovers a striking variation in the behavior of the x z and
yz on the bulk atom lying immediately below 4-fold coordinated CO. These levels do not interact with 2s* of
either the single species or the 4-fold ortho coadsorbed
system. However, nearly 10% of the x z and yz states are
pushed up into the 4-fold tetra 2s* peak. In simple terms,
an additional metal to CO backbonding interaction has
been “turned on”, thus the 2n* occupation increases. Why
is this particular interaction available at this site? It is
not due simply to the tetragonal symmetry, since the bulk
x z and yz do not interact with 2a* in a Ni(100)-c(2 x 2)CO
system. The latter has the same total adsorbate coverage
as the coadsorbed systems; each S is simply replaced by
an additional CO. We can speculate that the effect of the
S on the bulk atom below the 4-fold tetra CO is strong
enough to modify the levels in some way but too weak to
constrain their interactions.
Clearly, the mechanisms for interadsorbate interactions
are severely limited in this system. Not only does the
distance preclude a direct effect, but adsorbate-surface
interactions are segregated such that the potential strength
of an indirect, substrate-mediated effect is minimized as
well. At this point, a good explanation of the trends is still
obscure, and we hesitate even to speculate on its origins.
The effect of sulfur on CO adsorption appears to be
determined by the interadsorbate distance. At short
separations the S and CO, interaction is repulsive-we see
site blockage. At intermediate separations, for instance
in the energetically favorable geometry where S is in a
4-fold hollow and CO is bridging, we see the workings of
the usual Blyholder model. The 2s* of CO is populated
less, with resultant weakening of C-Ni and strengthening
of C-O bonds. We trace this effect to the modification by
S of the ability of specific surface orbitals ( x z and yz) to
back-bond to 2?r* or through-bond coupling. At longer
separations, we see a small reverse effect (C-0 weaker,
C-Ni stronger) which has some experimental support, but
it is easy to rationalize. In general, we think that the
simple electronegativity and electron-transfer explanation
of coadsorption effects needs to be supplemented by a
detailed orbital analysis in terms of through-space and
through-bond interactions.
Acknowledgment. We thank Susan Jansen for helpful
discussions during the course of this work. This work was
supported by the Office of Naval Research.
Appendix
The calculations were performed by using the tightbinding extended Huckel method.21aybThe Hi,%for Ni had
been previously determined28by charge iteration on the
bulk metal using Gray’s equations.39 The Hii’s for C and
0 had also been determined previously26by three-cycle
iteration on CO adsorbed onto a four-layer Fe(ll0) slab.
A, B, and C iteration parameters are from ref 38. All
parameters, including the uniterated S parameters, are
listed in Table VII.
For both the Ni(100)-p(2 X 2) S and 4 2 X 2) S systems,
the Ni-S bond length is reported to be 2.19
The
Ni(100)-p(2 X 2) CO geometry is d(Ni-C) = 1.80 i4 and
d(C-0) = 1.15 A.12c Adsorbates were placed on one side
of the three-layer slab. K point sets of 10 and 16 were used
respectively for systems of tetragonal and orthorhombic
symmetry. Convergence was checked for 10, 15, and 28
point sets on the Ni(100)-p(2 X 2) S system. The maximum variations among these sets were 0.007 eV for the
Fermi energy, 0.001 eV for the total energy, 0.010 for orbital electron densities, and 0.001 for overlap populations.
Registry No. N i , 7440-02-0; S, 7704-34-9; CO, 630-08-0.
(39) At d(Ni-Ni) = 2.49 A, the )r bonding interaction is too weak to
be detected in the COOP. These levels could interact with one of the CO
27r*.
(40) Ballhausen, C. J.; Gray, H. B. Molecular Orbital Theory; W. A.
Benjamin: New York, 1965; p 125,
(41) McGlynn, S. P.; Van Quickenborne, L. G.; Kinoshita, M.; Caroll,
D. G. Introduction to Applied Quantum Chemistry;Holt, Rinehart and
Winston: New York, 1972.